BPA: A Multilingual Sentiment Analysis Approach based on BiLSTM

: Sentiment analysis (SA) is the automatic process of understanding people’s feelings or beliefs expressed in texts such as emotions, opinions, attitudes, appraisals and others. The main task is to identify the polarity level (positive, neutral or negative) of a given text. This task has been the subject of several research competitions in many languages, for instance, English, Spanish and Arabic. However, developing a multilingual sentiment analysis method remains a challenge. In this paper, we propose a new approach, called BPA, based on BiLSTM neural networks, pooling operations and attention mechanism, which is able to automatically classify the polarity level of a text. We evaluated the BPA approach using five different data sets in three distinct languages: English, Spanish and Portuguese. Experimental results evidence the suitability of the proposed approach to multilingual and domain-independent polarity classification. BPA’s best results achieved an accuracy of 0.901, 0.865 and 0.923 for English, Spanish and Portuguese, respectively.

[1]  Jorge P'erez,et al.  Spanish Pre-trained BERT Model and Evaluation Data , 2023, ArXiv.

[2]  Roberto de Alencar Lotufo,et al.  BERTimbau: Pretrained BERT Models for Brazilian Portuguese , 2020, BRACIS.

[3]  Yi Cai,et al.  Recurrent neural network with pooling operation and attention mechanism for sentiment analysis: A multi-task learning approach , 2020, Knowl. Based Syst..

[4]  I. Maglogiannis,et al.  On the Reusability of Sentiment Analysis Datasets in Applications with Dissimilar Contexts , 2020, AIAI.

[5]  MATHEUS ARAUJO,et al.  A comparative study of machine translation for multilingual sentence-level sentiment analysis , 2020, Inf. Sci..

[6]  Hao Tian,et al.  ERNIE 2.0: A Continual Pre-training Framework for Language Understanding , 2019, AAAI.

[7]  Kai Zou,et al.  EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks , 2019, EMNLP.

[8]  Xing Wu,et al.  Conditional BERT Contextual Augmentation , 2018, ICCS.

[9]  Daniela Moctezuma,et al.  EvoMSA: A Multilingual Evolutionary Approach for Sentiment Analysis , 2018, IEEE Comput. Intell. Mag..

[10]  François Laviolette,et al.  Importance of Self-Attention for Sentiment Analysis , 2018, BlackboxNLP@EMNLP.

[11]  Siyuan Chen,et al.  A Deep Neural Network Model for Target-based Sentiment Analysis , 2018, 2018 International Joint Conference on Neural Networks (IJCNN).

[12]  Samuel R. Bowman,et al.  GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding , 2018, BlackboxNLP@EMNLP.

[13]  Christopher Potts,et al.  Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank , 2013, EMNLP.

[14]  Ming-Wei Chang,et al.  BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[15]  Feiyu Xu,et al.  PERSEUS: A Personalization Framework for Sentiment Categorization with Recurrent Neural Network , 2018, ICAART.

[16]  Gui Xiaolin,et al.  Deep Convolution Neural Networks for Twitter Sentiment Analysis , 2018, IEEE Access.

[17]  Daniele Bonadiman,et al.  Neural Sentiment Analysis for a Real-World Application , 2017, CLiC-it.